HNHacker News
TopNewBestAskShowJobs

galsapir

93 karma · joined January 17, 2026

sparse thoughts on things i find online
submissionscomments

Designing a domain-specific research harness

sparsethought.com·1 pts·galsapir·
0

Creative Reading: Scaffolding Reading for Transformation

arxiv.org·1 pts·galsapir·
0

No agents after 8 pm

sparsethought.com·2 pts·galsapir·
0

AI Has Plunged the Book Publishing Industry into Utter Chaos

wsj.com·23 pts·galsapir·
21

Can AI agents conduct open-ended AI research?

arxiv.org·4 pts·galsapir·
0

Ask HN: Any Good Pi.dev Setups?

3 pts·galsapir·
1

Giving a domain a hill to climb: benchmarking as data activation

sparsethought.com·12 pts·galsapir·
7

A bitter lesson for medicine, or a benchmark problem?

sparsethought.com·2 pts·galsapir·
0

Can LLMs Beat Classical Hyperparameter Optimization Algorithms?

arxiv.org·120 pts·galsapir·
20

Gemma 4 E4B as a primary local LLM (replaced Qwen)

digg.com·2 pts·galsapir·
0

PEEK: Give Your Agent an Orientation Cache (MIT CSAIL, Khattab group)

zhuohangu.github.io·3 pts·galsapir·
0

Hyperagents (Meta Research)

arxiv.org·2 pts·galsapir·
0

The Unreasonable Effectiveness of HTML

claude.com·3 pts·galsapir·
2

The Comparator in Clinical AI

sparsethought.com·2 pts·galsapir·
1

Borges' cartographers and the tacit skill of reading LM output

galsapir.github.io·40 pts·galsapir·
10

Best read of 2026 so far was written in 1880

galsapir.github.io·1 pts·galsapir·
1

Anthropic launched community ambassador program

claude.com·1 pts·galsapir·
0

LLMs as nudging research towards luke-warm middle

nature.com·1 pts·galsapir·
0

How do you evaluate a foundation model before you know what it's for?

galsapir.github.io·1 pts·galsapir·
1

Ask HN: Anyone using Claude Agent SDK in production?

1 pts·galsapir·
1

Data Activation Thoughts

galsapir.github.io·21 pts·galsapir·
5